Research map / Video / Video-language understanding
The SA-FARI Dataset: Segment Anything in Footage of Animals for Recognition and Identification
Dante Wasmuht, Otto Brookes, Maximilian Schall, Pablo Palencia, Christopher Beirne, Tilo Burghardt, Majid Mirmehdi, Hjalmar Kühl, Mimi Arandjelovic, Sam Pottie, Peter Bermant, Brandon Asheim and 12 more
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add SA-FARI data proportionally into SAM 3 training | SAM 3 · SA-FARI test (species-specific) | cgF1: 14.0 -> 39.0 | Improved |
| add SA-FARI data proportionally into SAM 3 training | SAM 3 · SA-FARI test (species-specific) | pHOTA Total: 48.5 -> 63.1 | Improved |
| explicitly fine-tune SAM 3 on SA-FARI | SAM 3 · SA-FARI test (species-specific) | cgF1: 14.0 -> 46.9 | Improved |
| explicitly fine-tune SAM 3 on SA-FARI | SAM 3 · SA-FARI test (species-specific) | pHOTA Total: 48.5 -> 68.1 | Improved |
| explicitly fine-tune SAM 3 on SA-FARI | SAM 3 · SA-FARI test (species-specific) | TETA: 39.6 -> 58.7 | Improved |
| fine-tune rather than mix SA-FARI into training | SAM 3 (SA-FARI) · SA-FARI test (species-specific) | cgF1: 39.0 -> 46.9 | Improved |
| use SAM 3 detector-tracker instead of LLMDet with tracker | LLMDet + SAM 3 TR · SA-FARI test (species-specific) | cgF1: 2.6 -> 14.0 | Improved |
| replace MegaDetector + tracker with SAM 3 trained on SA-FARI | MD+BoostSort++ · SA-FARI test (species-agnostic) | IDF1: 47.2 -> 71.1 | Improved |
| replace MegaDetector + OCSort with SAM 3 trained on SA-FARI | MD+OCSort · SA-FARI test (species-agnostic) | HOTA Total: 45.5 -> 64.4 | Improved |
| small instead of large masklets in test videos | SAM 3 FT (SA-FARI) · SA-FARI test subsets | pHOTA Total: 81.4 -> 52.2 | Hurt |
| challenging videos (occlusion, motion) vs overall set | SAM 3 FT (SA-FARI) · SA-FARI test subsets | pHOTA Ass: 84.6 -> 75.3 | Hurt |
| night videos vs overall set | SAM 3 FT (SA-FARI) · SA-FARI test subsets | pHOTA Total: 68.1 -> 66.0 | Hurt |
Closest papers by meaning
- MammAlps: A Multi-view Video Behavior Monitoring Dataset of Wild Mammals in the Swiss Alps (CVPR 2025 · similarity 0.91)
- BuckTales: A multi-UAV dataset for multi-object tracking and re-identification of wild antelopes (NeurIPS 2024 · similarity 0.90)
- The PanAf-FGBG Dataset: Understanding the Impact of Backgrounds in Wildlife Behaviour Recognition (CVPR 2025 · similarity 0.90)
- Generating the Wild: Individual-Consistent Image-to-Video Generation for Wildlife (NeurIPS 2026 · similarity 0.89)
- Web-Scale Collection of Video Data for 4D Animal Reconstruction (NeurIPS 2025 · similarity 0.88)
- BirdCollect: A Comprehensive Benchmark for Analyzing Dense Bird Flock Attributes (AAAI 2024 · similarity 0.88)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
